Reward-induced memory distortions bias value-based decisions

Our memories play a crucial role in many decisions. Recent work has begun to study how individuals recombine elements of past episodes stored in memory to evaluate the outcome of some decisions (Biderman et al., 2020; Shohamy & Daw, 2015). However, memory is not infallible and can get distorted. Understanding the impact of a fallible memory system on decision-making is essential. To explore how rewards influence memory and subsequent choices, we employed a modified version of the classic Deese-Roediger-McDermott (DRM) paradigm (Deese, 1959; Roediger & McDermott, 1995), where we assigned different levels of reward to different DRM lists at the encoding stage . Words were either novel, studied, or conceptually related critical lures. In Experiment 1, we tested the effect of reward level at encoding on subsequent memory performance. In Experiment 2, we tested the effect of reward level on subsequent memory-guided value-based decision making. We found that high-value critical lures were brought to mind more often than low-value critical lures overall. Interestingly, participants properly identified high-value critical lures as new, despite bringing those words to mind (i.e., participants successfully monitored their memory for high-value critical lures), but failed to so for low-value critical lures. Moreover, participants chose high-value critical lures more often than novel items that were not related to the studied lists when compared to low-value critical lures. These findings suggest that reward can modulate memory representations to be more generalizable and consequently shape decisions that rely on such representations.